- KDE developer Nate Graham’s draft LLM usage guidelines, requiring AI-generated code to be fully understood by its contributor and indistinguishable from human-written code, triggered a GitLab discussion that was locked and deleted on September 23, 2026 after two non-affiliated individuals were banned for derailing it.
- A “KDE for People” petition demanding a complete generative AI ban across all KDE projects drew signatures almost entirely from non-contributors, according to Nate Graham, undercutting its weight in a governance process where standing accrues to active maintainers.
Two people with no history in the KDE project hijacked its AI policy discussion, were exposed and banned, and the thread was deleted on September 23, 2026. The episode began with developer Nate Graham’s draft LLM usage guidelines and ended with a locked GitLab thread and a petition whose signatories were, by Graham’s own account, largely not KDE contributors at all.
Graham’s Draft Guidelines
The proposal sought a middle ground: AI tools are permitted, but the contributor remains fully responsible for the result. The core principle, which Graham framed as “don’t be lazy,” required that AI-generated code be indistinguishable from human-written code and that the contributor fully understand what they are submitting. The draft also banned AI-generated text for merge requests, commit messages and replies to review comments, on the basis that human-to-human communication matters in those contexts. The concern driving it was practical: maintainers across open-source projects have reported a rise in low-effort, AI-generated submissions that pass automated checks but burden reviewers. Early responses were broadly constructive, focused on refining wording rather than rejecting the premise.
How the Thread Collapsed
The discussion broke down when participants with no KDE history shifted the conversation from the specifics of the draft to broader ethical and environmental objections to AI. The situation worsened when one of those individuals publicly surfaced the controversial online history of another. Both were banned; the thread was deleted. Graham attributed the conflict to outside interference, describing the episode as an attempt to derail KDE’s effort to add restrictions to LLM usage rather than engage with its substance. The pattern is familiar in open-source governance debates over AI: a technical policy discussion, open by design, becomes a target for actors whose objection is to AI broadly, not to the specific project’s choices.
The Petition and Its Limits
Out of the collapsed thread came “KDE for People,” a group that launched a petition calling for a complete ban on generative AI across all KDE projects, covering code, assets, documentation, translations and discussion at community events including Akademy. The petition cited projects such as Zig, elementaryOS and OBS Studio as precedents and referenced an Amnesty International report on AI harms. Its framing was explicit: a “pro-people outlook to community, ethics and with respect for the work that has gone into KDE software and projects for decades.”
Graham noted that almost none of the signatories were recognised KDE contributors. In open-source governance, standing generally accrues to people who do the work; a petition signed primarily by non-contributors carries less procedural weight than one driven by active maintainers, regardless of its rhetorical force. The thread itself exposed the same problem: the people most vocal about KDE’s AI policy are largely not the people who bear the day-to-day cost of implementing it.
Where KDE Is Heading
The debate at KDE sits within a broader split in open-source communities over AI contributions. GIMP has implemented an outright ban on AI-generated content. Nobara Linux has moved in the opposite direction, actively adopting AI coding tools. The Linux kernel project occupies the middle ground that now appears most likely for KDE: AI-assisted code permitted provided it passes human review and meets existing quality standards, with contributors expected to demonstrate understanding of what they submit.
Fedora and Debian lean toward that pragmatic position, while Gentoo has taken a more restrictive stance and GNOME has adopted caution around submissions with significant AI-generated content. Whether Graham’s draft becomes KDE policy, and in what form, remains unresolved pending the community’s next governance step.
